On Mon, 7 Mar 2022 at 19:19, Daniel Henrique Barboza <danielhb...@gmail.com> wrote: > > Hi, > > I got a lot of noise trying to debug an AIX guest in a pseries machine when > running with > '-d unimp'. The reason is that there is no distinction between features > (in my case, hypercalls) that are unimplemented because we never considered, > versus features that we made a design choice not to implement. > > This series adds a new log type, LOG_UNSUPP, as a way to filter the > second case. After changing the log level of existing unsupported > pseries hypercalls, -d unimp was reporting just the ones that I need to > worry about and decide whether we should implement it or mark as > unsupported in our model. After this series there's still one hypercall > thgat is being thrown by AIX. We'll deal with that another day.
So the intention of the distinction is: LOG_UNIMP: we don't implement this, but we should LOG_UNSUPP: we don't implement this, and that's OK because it's optional ? I think I'd be happier about adding a new log category if we had some examples of where we should be using it other than just in the spapr hcall code, to indicate that it's a bit more broadly useful. If this is a distinction that only makes sense for that narrow use case, then as Philippe says a tracepoint might be a better choice. thanks -- PMM
